@asnd/skill-creator 0.1.2 → 0.1.3

package/prompts/skill-creator.md

@@ -0,0 +1,169 @@
+---
+description: Generate an API/tool skill and wrapper CLI
+argument-hint: '<source flags or URL> [--name skill-name] [--agent agent] [--scope project|global]'
+---
+
+# Task: create a reusable agent command surface
+
+Invocation arguments from the user/host:
+
+```txt
+$ARGUMENTS
+```
+
+Generate and install a complete Agent Skill plus thin wrapper CLI scripts from an MCP server, OpenAPI spec, GraphQL endpoint/schema, or similar API/tool source. This command is for creating reusable agent infrastructure, not for one-off API calls.
+
+If the arguments block is empty or still a literal placeholder, infer the source from the surrounding user message. If the source, target agent, or install scope is still missing, ask only the minimum blocking questions.
+
+## Research-backed command principles
+
+Follow these principles when creating the generated skill and its scripts:
+
+- Write instructions for the future agent, not prose about what the command would do.
+- Keep command/prompt descriptions short, verb-led, and specific.
+- Make the CLI inspectable: it must support discovery/help such as `--list`, `--search`, and `<command> --help`.
+- Make output small by default: document `--head`, `--limit`, `--fields`, `--pretty`, `jq`, and file export patterns when useful.
+- Preserve approval boundaries: identify write/delete/mutation commands and do not test destructive operations without explicit safe test data.
+- Avoid raw secrets. Use environment variables or files, e.g. `--auth-header 'Authorization:env:API_TOKEN'`.
+- Test the installed wrapper from the generated skill directory, exactly as a future agent will run it.
+
+## Decide the target
+
+1. Determine the target agent and install scope.
+   - Agents commonly supported by `npx @asnd/skill-creator generate`: `pi`, `universal`, `codex`, `claude-code`, `cursor`, `opencode`, `gemini-cli`, `amp`, `github-copilot`, `goose`, `cline`, `windsurf`.
+   - Scope is `project` or `global`.
+2. Choose a spec-compliant skill name: lowercase kebab-case, 1-64 chars, letters/numbers/hyphens only, no leading/trailing hyphen, no consecutive hyphens.
+3. Choose a script name, usually the same as the skill name.
+
+## Prefer the built-in generator
+
+Use `npx @asnd/skill-creator` for generator commands so the package does not need to be installed globally.
+
+### OpenAPI
+
+```bash
+npx @asnd/skill-creator generate \
+  --template openapi \
+  --name <skill-name> \
+  --spec <openapi-url-or-file> \
+  --agent <agent> \
+  --scope project|global
+```
+
+### GraphQL
+
+```bash
+npx @asnd/skill-creator generate \
+  --template graphql \
+  --name <skill-name> \
+  --graphql <graphql-endpoint-url> \
+  --graphql-schema <schema-url-or-file> \
+  --agent <agent> \
+  --scope project|global
+```
+
+If no schema is provided and introspection may fail, ask whether to introspect live or where to get SDL/introspection JSON.
+
+### MCP HTTP/SSE
+
+```bash
+npx @asnd/skill-creator generate \
+  --template mcp-http \
+  --name <skill-name> \
+  --mcp <mcp-url> \
+  --agent <agent> \
+  --scope project|global
+```
+
+### MCP stdio
+
+```bash
+npx @asnd/skill-creator generate \
+  --template mcp-stdio \
+  --name <skill-name> \
+  --mcp-stdio '<server-command>' \
+  --agent <agent> \
+  --scope project|global
+```
+
+Useful optional flags:
+
+```bash
+--script <script-name>
+--auth-header 'Header:env:ENV_NAME'
+--base-url <url>
+--description '<trigger-oriented generated skill description>'
+--force
+--no-test
+```
+
+## Refine the generated skill
+
+After generation, inspect `SKILL.md` and rewrite it if needed so future agents know when and how to use the wrapper.
+
+The generated skill should contain:
+
+- Trigger-oriented frontmatter `description` that names the service and when to use it.
+- Setup/auth environment variables; never raw secret values.
+- Usage commands:
+  - `./scripts/<script-name> --list`
+  - `./scripts/<script-name> --search '<topic>'`
+  - `./scripts/<script-name> <command> --help`
+  - `./scripts/<script-name> --pretty <command> <flags>`
+- Small-output guidance: limits, field selection, paging, raw/binary output, and `jq` examples when responses are nested.
+- Safety notes for write/delete/mutation operations.
+- References to bundled specs/schemas/docs instead of copied full help output.
+
+Expected generated layout:
+
+```txt
+<skill-name>/
+├── SKILL.md
+├── scripts/
+│   └── <script-name>
+├── references/
+│   ├── openapi-spec-MM-DD-YYYY.json|yaml       # OpenAPI when available
+│   └── graphql-schema-MM-DD-YYYY.graphql|json  # GraphQL when available
+└── assets/                                     # optional static resources
+```
+
+MCP skills usually do not need a source artifact in `references/`; add concise notes only when they help future agents.
+
+## Smoke test
+
+From the generated skill directory, run discovery/help first:
+
+```bash
+./scripts/<script-name> --list
+./scripts/<script-name> --search '<topic>'
+./scripts/<script-name> <command> --help
+```
+
+Then run at most one safe read-only command with bounded output, for example:
+
+```bash
+./scripts/<script-name> --pretty --head 3 <safe-read-command> <flags>
+```
+
+If auth, billing, network, or destructive side effects block testing, do not fake success. Record exactly what was skipped and why.
+
+## Manual path for unsupported sources
+
+If `npx @asnd/skill-creator generate` does not support the source:
+
+1. Create the selected agent/scope skill directory manually.
+2. Add executable scripts in `scripts/` that hide setup/source details and expose a simple, inspectable CLI.
+3. Add stable specs, schemas, docs excerpts, examples, or notes under `references/`.
+4. Write `SKILL.md` using the requirements above.
+5. `chmod +x scripts/<script-name>` and smoke-test from the skill directory.
+
+## Final response
+
+Report:
+
+- Installed skill path and target agent/scope.
+- Files created or changed.
+- Wrapper commands future agents should use.
+- Required environment variables or auth setup.
+- Commands tested and results.
+- Anything intentionally not tested.

package/dist/cli/package-SLCRJ4QY.js

@@ -1,89 +0,0 @@
-// package.json
-var package_default = {
-  name: "@asnd/skill-creator",
-  version: "0.1.2",
-  description: "Turn any MCP server, OpenAPI spec, or GraphQL endpoint into a CLI at runtime.",
-  keywords: [
-    "graphql",
-    "mcp",
-    "openapi",
-    "pi-package",
-    "skills"
-  ],
-  homepage: "https://github.com/sandiiarov/skill-creator#readme",
-  bugs: {
-    url: "https://github.com/sandiiarov/skill-creator/issues"
-  },
-  license: "MIT",
-  repository: {
-    type: "git",
-    url: "git+https://github.com/sandiiarov/skill-creator.git"
-  },
-  bin: {
-    "skill-creator": "dist/cli/main.js"
-  },
-  files: [
-    "dist",
-    "skills",
-    "prompts",
-    "LICENSE",
-    "README.md"
-  ],
-  type: "module",
-  scripts: {
-    build: "tsup",
-    dev: "tsx src/cli/main.ts",
-    fmt: "oxfmt",
-    "fmt:check": "oxfmt --check",
-    lint: "oxlint",
-    "lint:fix": "oxlint --fix",
-    test: "vitest run",
-    "test:watch": "vitest",
-    typecheck: "tsc --noEmit"
-  },
-  dependencies: {
-    "@apidevtools/json-schema-ref-parser": "^14.2.1",
-    "@graphql-tools/graphql-file-loader": "^8.1.14",
-    "@graphql-tools/json-file-loader": "^8.0.28",
-    "@graphql-tools/load": "^8.1.10",
-    "@graphql-tools/url-loader": "^9.1.2",
-    "@modelcontextprotocol/sdk": "^1.23.0",
-    ajv: "^8.20.0",
-    "ajv-formats": "^3.0.1",
-    cacache: "^21.0.0",
-    graphql: "^16.14.0",
-    "graphql-request": "^7.4.0",
-    ky: "^2.0.2",
-    minimatch: "^10.1.1",
-    "string-argv": "^0.3.2",
-    yaml: "^2.8.1",
-    zod: "^4.1.13"
-  },
-  devDependencies: {
-    "@types/cacache": "^20.0.1",
-    "@types/node": "^24.10.1",
-    execa: "^9.6.0",
-    oxfmt: "^0.51.0",
-    oxlint: "^1.66.0",
-    tsup: "^8.5.1",
-    tsx: "^4.20.6",
-    typescript: "^5.9.3",
-    vitest: "^4.0.14"
-  },
-  engines: {
-    node: ">=20"
-  },
-  packageManager: "pnpm@10.1.0",
-  pi: {
-    skills: [
-      "./skills"
-    ],
-    prompts: [
-      "./prompts"
-    ]
-  }
-};
-export {
-  package_default as default
-};
-//# sourceMappingURL=package-SLCRJ4QY.js.map

package/dist/cli/package-SLCRJ4QY.js.map

@@ -1 +0,0 @@
-{"version":3,"sources":["../../package.json"],"sourcesContent":["{\n  \"name\": \"@asnd/skill-creator\",\n  \"version\": \"0.1.2\",\n  \"description\": \"Turn any MCP server, OpenAPI spec, or GraphQL endpoint into a CLI at runtime.\",\n  \"keywords\": [\n    \"graphql\",\n    \"mcp\",\n    \"openapi\",\n    \"pi-package\",\n    \"skills\"\n  ],\n  \"homepage\": \"https://github.com/sandiiarov/skill-creator#readme\",\n  \"bugs\": {\n    \"url\": \"https://github.com/sandiiarov/skill-creator/issues\"\n  },\n  \"license\": \"MIT\",\n  \"repository\": {\n    \"type\": \"git\",\n    \"url\": \"git+https://github.com/sandiiarov/skill-creator.git\"\n  },\n  \"bin\": {\n    \"skill-creator\": \"dist/cli/main.js\"\n  },\n  \"files\": [\n    \"dist\",\n    \"skills\",\n    \"prompts\",\n    \"LICENSE\",\n    \"README.md\"\n  ],\n  \"type\": \"module\",\n  \"scripts\": {\n    \"build\": \"tsup\",\n    \"dev\": \"tsx src/cli/main.ts\",\n    \"fmt\": \"oxfmt\",\n    \"fmt:check\": \"oxfmt --check\",\n    \"lint\": \"oxlint\",\n    \"lint:fix\": \"oxlint --fix\",\n    \"test\": \"vitest run\",\n    \"test:watch\": \"vitest\",\n    \"typecheck\": \"tsc --noEmit\"\n  },\n  \"dependencies\": {\n    \"@apidevtools/json-schema-ref-parser\": \"^14.2.1\",\n    \"@graphql-tools/graphql-file-loader\": \"^8.1.14\",\n    \"@graphql-tools/json-file-loader\": \"^8.0.28\",\n    \"@graphql-tools/load\": \"^8.1.10\",\n    \"@graphql-tools/url-loader\": \"^9.1.2\",\n    \"@modelcontextprotocol/sdk\": \"^1.23.0\",\n    \"ajv\": \"^8.20.0\",\n    \"ajv-formats\": \"^3.0.1\",\n    \"cacache\": \"^21.0.0\",\n    \"graphql\": \"^16.14.0\",\n    \"graphql-request\": \"^7.4.0\",\n    \"ky\": \"^2.0.2\",\n    \"minimatch\": \"^10.1.1\",\n    \"string-argv\": \"^0.3.2\",\n    \"yaml\": \"^2.8.1\",\n    \"zod\": \"^4.1.13\"\n  },\n  \"devDependencies\": {\n    \"@types/cacache\": \"^20.0.1\",\n    \"@types/node\": \"^24.10.1\",\n    \"execa\": \"^9.6.0\",\n    \"oxfmt\": \"^0.51.0\",\n    \"oxlint\": \"^1.66.0\",\n    \"tsup\": \"^8.5.1\",\n    \"tsx\": \"^4.20.6\",\n    \"typescript\": \"^5.9.3\",\n    \"vitest\": \"^4.0.14\"\n  },\n  \"engines\": {\n    \"node\": \">=20\"\n  },\n  \"packageManager\": \"pnpm@10.1.0\",\n  \"pi\": {\n    \"skills\": [\n      \"./skills\"\n    ],\n    \"prompts\": [\n      \"./prompts\"\n    ]\n  }\n}\n"],"mappings":";AAAA;AAAA,EACE,MAAQ;AAAA,EACR,SAAW;AAAA,EACX,aAAe;AAAA,EACf,UAAY;AAAA,IACV;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF;AAAA,EACA,UAAY;AAAA,EACZ,MAAQ;AAAA,IACN,KAAO;AAAA,EACT;AAAA,EACA,SAAW;AAAA,EACX,YAAc;AAAA,IACZ,MAAQ;AAAA,IACR,KAAO;AAAA,EACT;AAAA,EACA,KAAO;AAAA,IACL,iBAAiB;AAAA,EACnB;AAAA,EACA,OAAS;AAAA,IACP;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,EACF;AAAA,EACA,MAAQ;AAAA,EACR,SAAW;AAAA,IACT,OAAS;AAAA,IACT,KAAO;AAAA,IACP,KAAO;AAAA,IACP,aAAa;AAAA,IACb,MAAQ;AAAA,IACR,YAAY;AAAA,IACZ,MAAQ;AAAA,IACR,cAAc;AAAA,IACd,WAAa;AAAA,EACf;AAAA,EACA,cAAgB;AAAA,IACd,uCAAuC;AAAA,IACvC,sCAAsC;AAAA,IACtC,mCAAmC;AAAA,IACnC,uBAAuB;AAAA,IACvB,6BAA6B;AAAA,IAC7B,6BAA6B;AAAA,IAC7B,KAAO;AAAA,IACP,eAAe;AAAA,IACf,SAAW;AAAA,IACX,SAAW;AAAA,IACX,mBAAmB;AAAA,IACnB,IAAM;AAAA,IACN,WAAa;AAAA,IACb,eAAe;AAAA,IACf,MAAQ;AAAA,IACR,KAAO;AAAA,EACT;AAAA,EACA,iBAAmB;AAAA,IACjB,kBAAkB;AAAA,IAClB,eAAe;AAAA,IACf,OAAS;AAAA,IACT,OAAS;AAAA,IACT,QAAU;AAAA,IACV,MAAQ;AAAA,IACR,KAAO;AAAA,IACP,YAAc;AAAA,IACd,QAAU;AAAA,EACZ;AAAA,EACA,SAAW;AAAA,IACT,MAAQ;AAAA,EACV;AAAA,EACA,gBAAkB;AAAA,EAClB,IAAM;AAAA,IACJ,QAAU;AAAA,MACR;AAAA,IACF;AAAA,IACA,SAAW;AAAA,MACT;AAAA,IACF;AAAA,EACF;AACF;","names":[]}

package/prompts/generate-skill.md

@@ -1,22 +0,0 @@
----
-description: Generate a new Agent/Pi skill from an MCP server, OpenAPI spec, or GraphQL endpoint
-argument-hint: '<--mcp URL|--mcp-stdio CMD|--spec URL|FILE|--graphql URL> [skill-name] [destination]'
----
-
-Generate a new Agent/Pi skill from this API source:
-
-```txt
-$ARGUMENTS
-```
-
-Use this workflow:
-
-1. If the source, desired skill name, auth requirements, or destination is unclear, ask only the minimum follow-up questions.
-2. Read and follow the `skill-creator` skill instructions. If the skill is not already loaded, load `/skill:skill-creator` or read the local `skills/skill-creator/SKILL.md` when present.
-3. Use `skill-creator` to discover commands with `--list` or `--search`, inspect important commands with `<command> --help`, and test representative read-only calls first.
-4. Create a focused `SKILL.md` for the target API.
-   - Default destination: `.pi/skills/<kebab-case-skill-name>/SKILL.md`.
-   - Use a user-provided destination if one is included in the arguments.
-5. Include only tested workflows, practical examples, auth/secret handling, pagination/field-selection notes, and important gotchas.
-6. Do not duplicate the entire CLI help output. Do not execute write/delete/destructive API operations unless the user explicitly requested them.
-7. Validate the generated skill structure and frontmatter before finishing.

package/skills/skill-creator/SKILL.md

@@ -1,260 +0,0 @@
----
-name: skill-creator
-description: Turn MCP servers, OpenAPI specs, and GraphQL endpoints into runtime CLIs. Use when a user wants to discover available tools/endpoints, list commands, inspect command parameters, call MCP/OpenAPI/GraphQL operations, test an API from the shell, or create a skill/workflow around an API without writing generated client code.
----
-
-# skill-creator
-
-Use `skill-creator` to turn an MCP server, OpenAPI spec, or GraphQL endpoint into a CLI at runtime. Commands and flags are discovered dynamically from the source.
-
-This TypeScript port currently supports:
-
-- OpenAPI specs from local files or remote URLs
-- MCP over stdio
-- MCP over Streamable HTTP with SSE fallback
-- GraphQL introspection, provided SDL schemas, and introspection JSON schemas
-- filtering, caching, auth headers, output formatting, and GraphQL stdin variables
-
-## Core workflow
-
-1. Connect to exactly one source.
-2. Discover available commands with `--list` or `--search`.
-3. Inspect a command with `<command> --help`.
-4. Execute the command with flags. Put global options before the subcommand and command-specific options after it.
-
-```bash
-# OpenAPI
-skill-creator --spec ./openapi.json --list
-skill-creator --spec ./openapi.json --pretty get-pet --id 1
-
-# MCP over HTTP: auto tries Streamable HTTP then SSE
-skill-creator --mcp https://mcp.example.com/mcp --list
-skill-creator --mcp https://mcp.example.com/mcp search --query "test"
-
-# MCP over stdio
-skill-creator --mcp-stdio "node server.js" --list
-skill-creator --mcp-stdio "node server.js" echo --message "hello"
-
-# GraphQL
-skill-creator --graphql https://api.example.com/graphql --list
-skill-creator --graphql https://api.example.com/graphql users --limit 10
-```
-
-## CLI reference
-
-```txt
-skill-creator [global options] <subcommand> [command options]
-
-Source (mutually exclusive, one required):
-  --spec URL|FILE       OpenAPI spec (JSON or YAML, local or remote)
-  --mcp URL             MCP server URL (HTTP/SSE)
-  --mcp-stdio CMD       MCP server command (stdio transport)
-  --graphql URL         GraphQL endpoint URL
-
-Options:
-  --auth-header K:V     HTTP header; values support env:NAME and file:/path
-  --transport TYPE      MCP HTTP transport: auto|streamable|sse (default: auto)
-  --base-url URL        Override OpenAPI base URL
-  --graphql-schema SRC  GraphQL SDL or introspection JSON schema FILE|URL
-  --cache-key KEY       Custom cache key
-  --cache-ttl SECONDS   Cache TTL (default: 3600)
-  --refresh             Bypass cache
-  --list                List available commands
-  --search PATTERN      Search by command name or description
-  --include GLOBS       Include command globs, comma-separated
-  --exclude GLOBS       Exclude command globs, comma-separated
-  --methods METHODS     OpenAPI method filter, e.g. GET,POST
-  --fields FIELDS       GraphQL selection override, e.g. "id name email"
-  --selection-depth N   GraphQL default selection depth (default: 2)
-  --stdin               Read GraphQL variables from stdin JSON
-  --pretty              Pretty-print JSON
-  --raw                 Print raw response body
-  --head N              Limit array output to first N records
-  --help, -h            Show help
-  --version             Show version
-```
-
-## Authentication
-
-Prefer `env:` or `file:` secret references. Do not pass raw tokens literally.
-
-```bash
-export API_TOKEN="Bearer ..."
-
-skill-creator --mcp https://api.example.com/mcp \
-  --auth-header Authorization:env:API_TOKEN \
-  --list
-
-skill-creator --spec https://api.example.com/openapi.json \
-  --auth-header x-api-key:file:/run/secrets/api-key \
-  list-items
-```
-
-## MCP patterns
-
-### Transport selection
-
-```bash
-# Default: Streamable HTTP, then SSE fallback
-skill-creator --mcp https://mcp.example.com/mcp --list
-
-# Force Streamable HTTP only
-skill-creator --mcp https://mcp.example.com/mcp --transport streamable --list
-
-# Force legacy SSE
-skill-creator --mcp https://mcp.example.com/sse --transport sse --list
-```
-
-### GitHub remote MCP
-
-```bash
-export GITHUB_MCP_PAT="Bearer $(gh auth token)"
-
-skill-creator \
-  --mcp https://api.githubcopilot.com/mcp/x/repos/readonly \
-  --auth-header Authorization:env:GITHUB_MCP_PAT \
-  --list
-
-skill-creator \
-  --mcp https://api.githubcopilot.com/mcp/x/repos/readonly \
-  --auth-header Authorization:env:GITHUB_MCP_PAT \
-  get-file-contents \
-  --owner github \
-  --repo github-mcp-server \
-  --path README.md
-```
-
-### Filesystem MCP over stdio
-
-Current npm installs may need `ajv` supplied explicitly:
-
-```bash
-skill-creator --mcp-stdio \
-  "npx -y -p ajv -p @modelcontextprotocol/server-filesystem mcp-server-filesystem /tmp" \
-  --list
-```
-
-On macOS, `/tmp` resolves to `/private/tmp`. If the filesystem server rejects `/tmp/foo`, retry with `/private/tmp/foo`.
-
-## OpenAPI patterns
-
-```bash
-# Discover operations
-skill-creator --spec https://petstore3.swagger.io/api/v3/openapi.json --list
-
-# Override base URL if the spec lacks usable servers
-skill-creator --spec ./openapi.json --base-url https://api.example.com/v1 --list
-
-# Filter to safer read-only operations
-skill-creator --spec ./openapi.json --methods GET --list
-
-# Include/exclude by generated command names
-skill-creator --spec ./openapi.json --include 'list-*' --exclude '*internal*' --list
-```
-
-## GraphQL patterns
-
-### Introspected endpoints
-
-```bash
-skill-creator --graphql https://beta.pokeapi.co/graphql/v1beta --list
-
-skill-creator --graphql https://beta.pokeapi.co/graphql/v1beta \
-  --fields "id name" \
-  pokemon-v2-pokemon \
-  --limit 3
-```
-
-### Endpoints without introspection
-
-If introspection is disabled, provide SDL or introspection JSON with `--graphql-schema`.
-
-```bash
-skill-creator --graphql https://api.example.com/graphql \
-  --graphql-schema ./schema.graphql \
-  --list
-
-skill-creator --graphql https://api.example.com/graphql \
-  --graphql-schema ./introspection.json \
-  users --limit 10
-```
-
-If no schema is provided and introspection fails, skill-creator tries a stale cached schema. If none exists, provide `--graphql-schema`.
-
-### Variables from stdin
-
-Use `--stdin` for GraphQL variables that are easier to pass as JSON.
-
-```bash
-echo '{"limit": 3}' | skill-creator \
-  --graphql https://beta.pokeapi.co/graphql/v1beta \
-  pokemon-v2-pokemon \
-  --stdin
-```
-
-### Selection sets
-
-Use `--fields` for precise output and to avoid oversized nested responses.
-
-```bash
-skill-creator --graphql https://beta.pokeapi.co/graphql/v1beta \
-  --fields "id name height weight" \
-  pokemon-v2-pokemon-by-pk \
-  --id 25
-```
-
-If no `--fields` is supplied, skill-creator builds a default selection set. Use `--selection-depth 1` to keep it shallow.
-
-## Caching and refresh
-
-Remote OpenAPI specs, MCP HTTP tool lists, and GraphQL schemas are cached under `~/.cache/skill-creator` by default. Override with `SKILL_CREATOR_CACHE_DIR`.
-
-```bash
-skill-creator --graphql https://api.example.com/graphql --refresh --list
-skill-creator --mcp https://mcp.example.com/mcp --cache-ttl 86400 --list
-```
-
-## Output handling
-
-```bash
-# Pretty JSON
-skill-creator --spec ./openapi.json --pretty list-pets
-
-# Raw response body
-skill-creator --spec ./openapi.json --raw download-report > report.json
-
-# Preview large arrays
-skill-creator --graphql https://api.example.com/graphql --head 5 users
-```
-
-## Creating another skill around an API
-
-When asked to create a skill for a specific API, use skill-creator to explore first, then document the practical findings.
-
-1. Discover commands:
-
-   ```bash
-   skill-creator --mcp https://target.example.com/mcp --list
-   skill-creator --spec https://target.example.com/openapi.json --list
-   skill-creator --graphql https://target.example.com/graphql --list
-   ```
-
-2. Inspect important commands:
-
-   ```bash
-   skill-creator --mcp https://target.example.com/mcp <command> --help
-   ```
-
-3. Test representative read-only calls first. Use `--head`, `--fields`, and `--pretty` to control output size.
-
-4. Record gotchas in the generated `SKILL.md`:
-   - required auth scopes and safe secret passing
-   - pagination defaults and limits
-   - date/time formats
-   - fields that cause huge responses
-   - binary or raw output handling
-   - write/delete operations to avoid unless explicitly requested
-
-5. Prefer wrapper scripts or documented command templates for repeated connection flags. This TypeScript port does not include bake mode.
-
-Do not duplicate the entire `--help` output in generated skills. Focus on tested workflows, surprising defaults, anti-patterns, and examples that are known to work.